See discussions, stats, and author profiles for this publication at: [Link]

net/publication/272910775

Modeling and Simulation of Three-Phase Voltage Source Inverter Using a

Model Predictive Current Control

Article in International Journal of Innovation Management and Technology · January 2014

DOI: 10.7763/IJIMT.2014.V5.477

CITATIONS READS

31 1,865

1 author:

Ali Almaktoof
Cape Peninsula University of Technology
21 PUBLICATIONS 54 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Multi-level inverters using finite-set model predictive current control for Renewable Energy Systems applications View project

[Link]/IET Student Conference 22 October 2019 View project

All content following this page was uploaded by Ali Almaktoof on 01 June 2020.

The user has requested enhancement of the downloaded file.

 International Journal of Innovation, Management and Technology, Vol. 5, No. 1, February 2014

Modeling and Simulation of Three-Phase Voltage Source

Inverter Using a Model Predictive Current Control
Ali M. Almaktoof, A. K. Raji, and M. T. E. Kahn

 control method is known as a finite state model predictive

Abstract—This paper focuses on a combination of three- control (FS-MPC) since the possible control actions
phase Voltage Source Inverter (VSI) with a predictive current (switching states) are finite and this control strategy is used
control to provide an optimized system for three-phase in this paper. It has been successfully applied to a wide
inverter that controls the load current. The present range of power converter and drive applications [2], in a
contribution presents a Finite Set-Model Predictive Control three-phase inverter [4]–[7] and a matrix converter [8] and
(FS-MPC) strategy for a two-level three-phase VSI with
resistive- inductive load (RL-Load). In order to reduce the flux and torque control of an induction machine [9]. An
computational effort which gives rise to multiple possibilities example of different controlled variables using a single cost
has been determined. With different cases the evaluation of the function is presented in [10], where the current is controlled
system is done. Firstly, the system performance with long while, at the same time, minimizing the switching frequency
prediction horizon is carried out. Secondly, the dynamic and balancing the dc-link voltages in an inverter. In all these
response of the system with step change in the amplitude of the works the switching states are changed at equidistant time
reference is investigated. Simulations are carried out using
MATLAB/Simulink to test the effectiveness of an FS-MPC for instants.
the two-level VSI with RL-Load. In this paper a combination of three-phase VSI with a
predictive current control to provide an optimized system
Index Terms—Finite Set-Model predictive control, voltage for three-phase inverter that controls the load current. The
source inverter, RL-Load, prediction horizon. powerful and robustness of the proposed control method are
evaluated through simulations results.
Section two presents the power converter model that is
I. INTRODUCTION
used in this paper. Firstly the time-continuous model is
Voltage source converters have been extensively studied presented and then discretized. The control scheme is
in the last decades in most industrial sectors for many developed in section three that is used to control the power
applications. By considering the increasing energy demands converter model in section two. Matlab/Simulink modeling
and power quality and efficiency; a control and power and simulation work is presented in the penultimate section.
conversion using power electronics have become an Conclusion is presented in the last section of the paper.
important topic today.
Predictive control for power electronics has been
presented since 1980's [1]. Several advantages make the II. POWER CONVERTER MODEL
predictive control suitable for the control of power A. Voltage Source Inverter Layout
converters, for instance; easy to understand and implement
A two-level VSI three-phase power converter is the least
and can be applied to different kinds of voltage source
complicated multiple levels VSI because it presents only
converters. It requires a high amount of calculations,
two voltage levels. It has been selected for a clear analysis
compared to classic control methods, but the fast
of a predictive control strategy with RL-Load. The topology
microprocessors available today have made it possible to
of the inverter considered in this paper is depicted in Fig.1.
implement predictive control for VSI converters, and model
The circuit is operated by switching S1, S2, S3, S4, S5 and S6.
predictive control (MPC) have distinct advantages when
The inverter uses two pairs of complementary controlled
compared to the traditional PWM methods [2], [3].
switches in each inverter phase or leg, (S1, S2), (S3, S4) and
MPC for power converters and drives is to take
(S5, S6) as shown in Fig. 1. Considering that the two
advantage of the inherent discrete nature of power
switches in each inverter phase or leg operate in a
converters. Since power converters have a finite number of
complementary pair in order to avoid short circuiting the
switching states, the MPC optimization problem can be
DC source.
simplified and reduced to the prediction of the system
behavior only for those possible switching states. Then,
each prediction is used to evaluate a cost function (also S1 S3 S5
known as quality or decision function), and consequently RL-Load
i load
the switching state with minimum cost is selected, a
VDC n
generated and applied in the next switching instant. This b
c
RL-Load

RL-Load
S2 S4 S6
Manuscript received on August 28, 2013; revised November 9, 2013.
A. M. Almaktoof, R. K. Atanda, and T. Kahn are with Cape Peninsula
University of Technology. Electrical Engineering department, Bellville
Campus, 7535, Republic of South Africa (e-mail: Fig. 1. Two-level voltage source inverter circuit topology.
213301091@[Link], rajia@[Link], KhanT@[Link]).

DOI: 10.7763/IJIMT.2014.V5.477 9
 International Journal of Innovation, Management and Technology, Vol. 5, No. 1, February 2014

The state of the switches is determined according to:

(4)

A discrete-time equation for the future load current is

obtained by using Euler-forward equation as in (5) in order
to obtain a discrete-time system representation. The
derivation of the state variables is approximated as follows:

This leads to eight different switching possibilities and (5)

consequently, eight voltage vectors are obtained that is
voltages VDC and - VDC can be switched. As shown in Fig. 2 Ts is the sampling time and k is the current sampling
the voltage vectors which are generated by the inverter instant. The state variable is denoted with x. Then, the
resulted in only seven different voltage vectors because V0 discrete-time load model can be to:
and V7 produce the same zero voltage vector (V0 = V7),
that means a three-phase two-level voltage source converter
can deliver only 7 different voltage vectors, although there
are 8 different switching combinations, as it can be seen in
Fig. 2. The switching states that define the value of the
output voltages are determined according to:
(6)
I
V3 m V2

010 110
Equation (6) is used to predict the load current for each
switching possibility. The cost function g is evaluated for
V7 each of the seven possible voltage vectors generated by this
V4 V1 inverter to calculate the future value of the load current. The
111 R
011 100
000 e voltage vector that minimizes the cost function is selected
V0
and applied during the next sampling instant.

V5 001 101 V6 III. CONTROL STRATEGY

vaN = Sa VDC A. Finite Set-Model Predictive Control
vbN = Sb VDC An FS-MPC does not need use a modulator. FS-MPC has
vcN = Sc VDC been used as a current controller [11], [12] for two- [2]-[4],
three- [13] and four- [14], [15], level inverters. Fig. 3 shows
Fig. 2. Voltage vectors generated by 2-level VSI.
VSI converter under FS-MPCC, where: iref represents the
reference current for the predictive current control, i(k) is
B. Load Model the m measurements taken at time k and i(k+1) are the
The differential equation of the load current is applied to predicted values of the m states for n possible switching
obtain the continuous-time state-space equations of the load states at time k+1. The error between the reference and
for each phase: predicted values is obtained to minimize the cost function
and the switching state that minimizes the cost function is
VDC (t) = R . i(t) + L (1) chosen. The chosen state's switching signals, S, are then
applied to the converter.
Using the Clarke transformation, the equations can be i ref

expressed in the stationary α-β frame. Minimization of the

cost function
S RL-
Load
Clarke transformation is defined as following, i(k)
MPC
i(k+1)
n
2-Level VSI

Vα = 2/3 (Va – 0.5 Vb – 0.5 Vc) (2)

i(k) measurements

Vβ = 2/3 (0.5 Vb – 0.5 c) (3)

Fig. 3. FS-MPC block diagram.

Applying Clarke transformation on the load current, the In order to reduce the computational effort which gives
equations can be expressed in the stationary α-β frame as in rise to multiple possibilities (8 different switching
(4). Then, the continuous-time state-space equation of the possibilities for one prediction step) the best one of the 7
load results to: voltage vectors is determined; after that the best switching

10
 International Journal of Innovation, Management and Technology, Vol. 5, No. 1, February 2014

state which delivers this voltage vector is determined. In

order to determine the best one of the 7 voltage vectors that (11)
should be applied in the next sampling cycle k+1, the 1-
norm cost function has to be minimized, for the current
control of a 2-level VSI, a simple cost function can be
defined in absolute value term with one prediction step and When an inverter is controlled directly it gives rise to
n prediction steps respectively as: multiple possibilities. In this case the inverter leads to 8
different switching possibilities and consequently, eight
voltage vectors are obtained for one prediction step, with
(7) long prediction horizon the calculation effort rises, for two
prediction steps 64 possibilities, for three prediction steps
512 possibilities. With long prediction horizon the
possibilities of a VSI have been evaluated for RL-Load
(8)
using MATLAB/Simulink. Fig. 4 shows number of possible
switching states with long prediction steps for three-phase
In order to simplify the calculations, it can be assumed two-level VSI.
that the current reference values is constant in equation (9)
with the prediction horizon for small sampling time TS as in
equation (10). This approximation is considered in this
paper.

(9)

For more accurate approximation the reference current

values for α-β axis required by (7) have to be
predicted from the present and previous values of the Fig. 4. Number of possible switching states with long prediction steps.
current reference using a second-order extrapolation given
by
IV. SIMULATION RESULTS
(10)
FS-MPC strategy for three-phase two-level VSI has been
simulated with MATLAB/Simulink, in order to evaluate the
Equation (10) has been obtained from the quadratic
performance of the proposed control algorithm and check
Lagrange extrapolation formula [3], [16].
the performance and robustness of the proposed predictive
In general, the control algorithm as shown in Fig. 3 can
control method. A sinusoidal reference current was applied
be summarized to the following steps [17]:
to the system and the amplitude of the reference current was
1) Measure the load currents.
set to 4 A and frequency of 50 Hz per phase, an RL-Load is
2) Predict the load currents for the next sampling instant
connected to the output of the VSI as shown in Fig. 4. Table
for all the possible switching states.
I. Shows the parameters used for the simulations.
3) Evaluate the cost function for each prediction.
4) Optimal switching state is selected which minimizes TABLE I: PARAMETERS USED FOR THE SIMULATIONS
the cost function. Parameter Value
Load resistor, R 10 Ω
5) Apply the new switching state. Load inductor, L 10 mH
In the current control case, the cost function is minimized AC Filter, L 10 mH
which obtained from the error between the reference current DC link voltage, VDC 100 V
and predicted current for produce the switching state and Reference amplitude current, iref 4A
Sampling time, Ts 25μs , 75µs
applied.
B. Prediction Step
From i(k+1) on the controller can be seen in Fig. 3 can
use any possible output to bring the controlled currents that
closer to their reference. In the next step i(k+2), the
prediction step, the controller thus calculates the measured
currents at k+2 for all possible switch states at k+1. With
long prediction horizon (n steps) the output current with
RL-Load can be expressed as

Fig. 4. Modeling an FS-MPC of VSI using MATLAB /simulink.

11
 International Journal of Innovation, Management and Technology, Vol. 5, No. 1, February 2014

C. With long Prediction Horizon the proposed control algorithm to track sinusoidal reference
The robustness of the proposed control method was currents. It can be observed from this simulation that the
tested with four prediction steps, it can be seen in Fig. 5 predictive control method has fast dynamic response with
how the output currents are tracking their references with inherent decoupling between iα and iβ. It is again, the
four prediction steps, it can be seen that the control algorithm shows excellent tracking behavior.
algorithm shows excellent tracking behaviour. 6

The error between the reference and actual current for 4

different values of prediction horizon has been evaluated to
2
check the influence of the prediction horizon as shown in

iL & iref [A]

Fig. 6. 0

6 6

-2
4 4

-4
[A]
[A]

2 2
L
L

& i , i
& i , i
L

0 0
-6
ref

0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07
ref

i , i
i , i

ref
ref

-2 -2 Time [s]
-4 -4
(a)

-6
80
-6
0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08
Time [s] Time [s]
60
(a) (b)
6 6 40

4 4 20
[A]

v [V]
2
[A]

2
L

0
& i  , i
L
& i , i

L
L

0 0 -20
ref

ref
i , i

, i
ref

-2 -2
-40
ref
i

-4 -4
-60

-6 -6
0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08 -80
Time [s] 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07
Time [s]
(c) (d) Time [s]
(b)
Fig. 5. Simulations results: the load current with RL-Load for different
values of prediction horizon for a sampling time TS = 25 µs: Fig. 7. Simulations results: Sinusoidal reference steps: (a) Load current
a) n = 1. b) n = 2. c) n = 3. d) n = 4. with two step changes. (b) Load voltage vα changes during the reference
steps.
1 1

With constant reference steps: in this simulation the

0.5 0.5
control algorithm was tested with constant reference steps.
Error [A]

Error [A]

0 0
The simulation shows the control result for constant
reference steps values, when a step change in the amplitude
-0.5 -0.5
of the reference iα and iβ were changed from 0 A to 4 A and
-1 -1
0 A to -3 A respectively at 0.02 s as shown in Fig. 8. It can
0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08
Time [s] Time [s] be seen in Fig. 8 how the load currents reach to their
(a) (b)
1 1
reference with fast dynamic response.
This simulation clearly demonstrates that the proposed
0.5 0.5 control algorithm is able to track the currents to their
reference values. The remaining current ripple in steady
Error [A]
Error [A]

0 0
state results from the finite switching and controller
-0.5 -0.5
frequency.
6
-1 -1
0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08
Time [s] Time [s] 4
(c) (d)

2 4.5
Fig. 6. Simulations results: Error between the reference and load current
iL & iref [A]

for different values of prediction horizon for a sampling time TS = 25µs: 4

a) n = 1. b) n = 2. c) n = 3. d) n = 4. 0
3.5
0.025 0.03 0.035 0.04
-2
D. With Step Change
-4
With sinusoidal reference steps: The second simulation
shows the control result for sinusoidal reference values. The -6
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
result for step changes in the amplitude of the references iα Time [s]
and iβ were changed from 4 A to 2 A and 4 A to 3 A at
Fig. 8. Simulations results: Constant reference steps for load current.
0.042 s and 0.048 s respectively are shown in Fig. 7 (a), it
can be observed in Fig. 7 (b) for the same test that during
the step change of the currents iα and iβ the load voltage vα is
V. CONCLUSIONS
kept at its maximum value until the reference current iα is
achieved. This simulation clearly demonstrates the ability of In this paper the FS-MPC for two-level voltage source

12
 International Journal of Innovation, Management and Technology, Vol. 5, No. 1, February 2014

inverters has been presented. The proposed control does not Output LC Filter for UPS Applications,‖ IEEE Africon 2011 - The
Falls Resort and Conference Centre, Sep. 2011.
need to use modulator. The control algorithm has been [14] G. Perantzakis, F. Xepapas, and S. Manias, ―Efficient predictive
evaluated with three different cases through simulation current control technique for multilevel voltage source inverters,‖
results. First of all, the robustness of the proposed control presented at European Conference on Power Electronics and
Applications, Sep. 2005.
method with four prediction steps has been assessed; the [15] F. Defay, A. M. Llor, and M. Fadel, ―Direct Control Strategy for a
assessment has been done by checking the load current and Four-Level Three-Phase Flying-Capacitor Inverter,‖ IEEE
the error between the reference and actual current for four Transactions on Industrial Electronics, vol. 57, no. 7, 2010, pp.
prediction steps. It has been noticed that the control 2240–2248.
[16] O. Kukrer, ―Discrete-time current control of voltage-fed three-phase
algorithm provides very good current tracking behaviour. PWM inverters,‖ IEEE Trans. Ind. Electron., vol. 11, no. 2, Mar.
Secondly, with step change in the amplitude of the reference, 1996, pp. 260–269.
the simulation results shows that the predictive control [17] P. Cortes, J. Rodriguez, C. Silva, and A. Flores, ―Delay compensation
in model predictive current control of a three-phase inverter,‖ IEEE
method has fast dynamic response with inherent decoupling
Transactions on Industrial Electronics, vol 59, no. 2, 2012, pp.
between iα and iβ. Simulation results show that FS-MPC 1323–1325.
strategy gives very good performance under these
conditions. Ali M. Almaktoof was born in Libya in 1978. He
received the B. Eng. from Al-Tahady University,
Libya in 2001 and [Link]. from University of Tripoli,
REFERENCES Libya in 2008, all in electrical engineering.
[1] R. Kennel and D. Schroeder, Predictive control strategy for He is currently working toward the Doctor of
converters, in Proc. the third IFAC Symposium, pp. 415–422, Electrical Technology (DTech) degree in electrical
Lausanne, 1983. engineering at Cape Peninsula University of
[2] J. Rodrí guez and C. Cortés, Predictive control of Power Converters Technology (CPUT) in South Africa.
and Electrical Drives, UK: A Johan Wiley & Sons, Ltd., 2012. He is a member of World Society of Sustainable Energy Technology
[3] J. Rodrí guez, J. Pontt, C. A. Silva, P. Correa, P. Lezana, P. Cortes, (WSSET). He is an active member of the Centre for Distributed Power and
and U. Ammann, ―Predictive current control of a voltage source Electronics Systems at CPUT.
inverter,‖ IEEE Transactions on Industrial Electronics, vol. 54, no. 1, Mr. Almaktoof research interests include control of power converters,
February 2007, pp. 495–503. model predictive control, multilevel converters and renewable energy
[4] H. Abu-Rub, J. Guzinski, Z. Krzeminski, and H. A. Toliyat, systems applications.
―Predictive current control of voltage source inverters,‖ IEEE
Transactions on Industrial Electronics, vol. 51, no. 3, Jun. 2004, pp. Raji K. Atanda was born in Nigeria in 1969. He
585–593. received the [Link] (Electrical) from University of
[5] J. Li, F. Zhuo, X. Wang, L. Wang, and S. Ni, ―A grid-connected PV Ilorin in 1992; MTech (Electrical) and [Link].
system with power quality improvement based on boost + dual-level
(Electronics) from Cape Peninsula University of
fourleg inverter,‖ in Proc. IEEE Int. Power Electron. and Motion
Technology (CPUT) and Ecole Supérieure
Control Conf., Wuhan, China, May 2009, pp. 436–440.
[6] V. Yaramasu, M. Rivera, B. Wu, and J. Rodriguez, ―Model d'Ingénieurs en Electronique et Electrotechnique
Predictive Current Control of Two-Level Four-Leg Inverters—Part I: France in 2009 respectively and DTech (Electrical)
Concept, Algorithm, and Simulation Analysis,‖ IEEE Transactions from CPUT in 2013.
on Industrial Electronics, vol. 28, no. 7, July 2013, pp. 3459 – 3468. He is currently a lecturer and researcher at the department of electrical
[7] R. Vargas et al., ―Predictive control of a three-phase neutral point in CPUT and also the programme coordinator of the BTech Electrical
clamped inverter,‖ IEEE Trans. Ind. Electron., vol. 54, no. 5, pp. programme. He is a member of Power Electronics society of IEEE, Power
2697–2705, Oct. 2007. System society of IEEE and National Energy Association of South Africa
[8] S. Muller, U. Ammann, and S. Rees, ―New time-discrete modulation and an active member of the Centre for Distributed Power and Electronics
scheme for matrix converters,‖ IEEE Trans. Ind. Electron., vol. 52,
no. 6, Dec. 2005, pp. 1607–1615. Dr. Raji research interest is in the application of power electronics
[9] J. Rodrí guez, J. Pontt, C. Silva, P. Cortés, S. Rees, and U. Ammann, technology and control system development for alternative electricity
―Predictive direct torque control of an induction machine,‖ presented generation, transmission, distribution and utilization.
at Power Electron. Motion Control Conf., Sep. 2–4, 2004, Riga,
Latvia, Tariq Kahn was born in South Africa. He holds a
[10] R. Vargas, U. Ammann, J. Rodrí guez, and J. Pontt, ―Reduction of doctoral degree in electrical engineering from Cape
switching losses and increase in efficiency of power converters using Peninsula University of Technology.
predictive control,‖ in Proc. IEEE PESC, Jun. 15–19, 2008, pp. He is currently the head of the Center for
1062–1068. Distributed Power and Electronic Systems at the
[11] E. Camacho and C. Bordons, Model Predictive Control, NY:
department of Electrical Engineering in CPUT.
Springer-Verlag, 1999.
Professor Kahn research interests include: optics,
[12] P. Cortes et al., ―Predictive control in power electronics and drives,‖
MEMS, green energy technology, energy efficiency, distributed
IEEE Transactions on Industrial Electronics, vol. 55, no. 12, Dec.
2008, pp. 4312–4324. generation, automation and power electronics technology.
[13] P. Stolze, D. du Toit, M. Tomlinson, R. Kennel, and T. Mouton,
―Model Predictive Control of a Flying Capacitor Converter with

13

View publication stats